mirror of
https://gitlab.com/libeigen/eigen.git
synced 2025-03-07 18:27:40 +08:00
Fix my recent prefetch changes:
- the first prefetch is actually harmful on Haswell with FMA, but it is the most beneficial on ARM. - the second prefetch... I was very stupid and multiplied by sizeof(scalar) and offset of a scalar* pointer. The old offset was 64; pk = 8, so 64=pk*8. So this effectively restores the older offset. Actually, there were two prefetches here, one with offset 48 and one with offset 64. I could not confirm any benefit from this strange 48 offset on either the haswell or my ARM device.
This commit is contained in:
parent
488874781b
commit
26275b250a
@ -855,7 +855,7 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
|
||||
EIGEN_ASM_COMMENT("begin step of gebp micro kernel 3pX4"); \
|
||||
EIGEN_ASM_COMMENT("Note: these asm comments work around bug 935!"); \
|
||||
internal::prefetch(blA+(3*K+16)*LhsProgress); \
|
||||
internal::prefetch(blB+(4*K+16)*RhsProgress); /* Bug 953 */ \
|
||||
if (EIGEN_ARCH_ARM) internal::prefetch(blB+(4*K+16)*RhsProgress); /* Bug 953 */ \
|
||||
traits.loadLhs(&blA[(0+3*K)*LhsProgress], A0); \
|
||||
traits.loadLhs(&blA[(1+3*K)*LhsProgress], A1); \
|
||||
traits.loadLhs(&blA[(2+3*K)*LhsProgress], A2); \
|
||||
@ -878,7 +878,7 @@ void gebp_kernel<LhsScalar,RhsScalar,Index,DataMapper,mr,nr,ConjugateLhs,Conjuga
|
||||
EIGEN_ASM_COMMENT("end step of gebp micro kernel 3pX4"); \
|
||||
} while(false)
|
||||
|
||||
internal::prefetch(blB + 4 * pk * sizeof(RhsScalar)); /* Bug 953 */
|
||||
internal::prefetch(blB + 8 * pk); /* Bug 953 */
|
||||
EIGEN_GEBP_ONESTEP(0);
|
||||
EIGEN_GEBP_ONESTEP(1);
|
||||
EIGEN_GEBP_ONESTEP(2);
|
||||
|
Loading…
Reference in New Issue
Block a user